Marine Ecology Progress Series 333:37

MARINE ECOLOGY PROGRESS SERIES Vol. 333: 37–50, 2007 Published March 12 Mar Ecol Prog Ser

Food web interactions along seagrass–coral reef boundaries: effects of piscivore reductions on cross-habitat energy exchange

John F. Valentine1, 2,*, Kenneth L. Heck Jr.1, 2, Derrick Blackmon1, 2, Margene E. Goecker1, 2, Juliet Christian1, 2, Ryan M. Kroutil1, 2, Kevin D. Kirsch3, Bradley J. Peterson4, Mike Beck5, Mathew A. Vanderklift6, 7

1Dauphin Island Sea Laboratory, University of South Alabama, 101 Bienville Boulevard, Dauphin Island, Alabama 36528-0369, USA 2Department of Marine Science, University of South Alabama, Mobile, Alabama 36688, USA 3NOAA Damage Assessment Center, PO Box 500368, Marathon, Florida 33050, USA 4Marine Science Division, Southampton College, Southampton, New York 11968-4198, USA 5The Nature Conservancy, 100 Shaffer Road, University of California, Santa Cruz, California 95060, USA 6CSIRO Marine Research, Underwood Avenue, Floreat, Western Australia, Australia 7Centre for Ecosystem Management, Edith Cowan University, 100 Joondalup Drive, Joondalup, Western Australia, Australia

ABSTRACT: Although early studies qualitatively documented the importance of cross-habitat energy transfers from seagrasses to coral reefs, such exchanges have yet to be quantified. Empirical evidence suggests that grazing by reef-associated herbivores along the coral reef–seagrass interface can be intense (e.g. conspicuous presence of bare-sand ‘halos’ surrounding coral reefs). This evidence must be interpreted with caution, however, as most of it comes from areas that have experienced sus- tained, intense overfishing. To quantify the impacts of piscivore removal on cross-habitat energy exchange at the coral reef–seagrass interface, we compared grazing intensity along fished and no- take reefs in the upper and lower Florida Keys. Using visual census techniques and direct measures of seagrass grazing, we documented the impacts of piscivore density on herbivory along the seagrass–coral reef interface. Grazing rates were greater than observed seagrass (Thalassia testudinum) production near reefs in the upper Keys, but less than 48% of production in the lower Keys. Analyses showed that these large differences were not related to regional differences in either herbivore density or species composition. Seagrass biomass was also lower near reefs in the upper Keys, where estimates of grazing were highest. Piscivores were dominated by transient predators (jacks and barracudas) whose densities varied with region and distance from reef, but not with protection from fishing. A nonsignificant negative correlation between great barracuda density and leaf losses from tethered shoots, coupled with the greater abundances of larger grazers near reefs, suggests that predation risk, rather than direct reductions in density, may limit grazers to intense feeding on seagrasses adjacent to reefs in the upper Keys. The large-scale variation in grazing intensity illustrates the need for more detailed quantifications of energy exchanges along the seagrass–coral reef boundary.

KEY WORDS: Edge effects · Herbivory · Overfishing · Parrotfish · Spatial subsidies · Trophic transfer

Resale or republication not permitted without written consent of the publisher

INTRODUCTION either low or unavailable to higher-order consumers (Polis & Strong 1996). In some cases, these exchanges Cross-habitat exchanges of nutrients and energy (i.e. are so strong that they can allow consumers to exist at spatial subsidies) can be of great importance, espe- much higher densities than would be possible if they cially in areas where in situ primary production is relied solely on autochthonous production (e.g. Hilder-

*Email: [email protected] © Inter-Research 2007 · www.int-res.com 38 Mar Ecol Prog Ser 333: 37–50, 2007

brand et al. 1999, Nakano & Murakami 2001). While It is also noteworthy that previous descriptions of ecologists working at the terrestrial–aquatic interface seagrass herbivory were based primarily on observa- have quantified the strength of such exchanges (Pow- tions from sites now known to have been intensely ers 2001), few have done so for marine environments fished long before modern day investigations were (Cadenasso et al. 2003). conducted (Pandolfi et al. 2003). Among the many Despite recent investigations of the importance of impacts of overfishing has been a worldwide depletion exogenous inputs of nutrients to coral reef food webs of piscivorous fish biomass to around 10% of historical (Leichter et al. 2003), most researchers adhere to the levels (Myers & Worm 2003). Additionally, herbivore view that reef communities are closed entities charac- biomass has been reduced on a number of reefs terized by tight recycling processes (Hatcher 1997). through indiscriminate harvesting by fishers using Although early evidence qualitatively documented the traps (e.g. Mumby et al. 2006). The consequences of importance of cross-habitat energy transfer from sea- these reductions in fish density for trophic exchanges grasses to coral reefs (see review by Valentine & Duffy between seagrass and coral reefs are unknown, but it 2006), such exchanges have yet to be quantified. seems clear that they have been reduced (Valentine & Instead, most studies rely on descriptions of poorly Heck 2005). Conversely, it is also possible that recent resolved food webs (Link 2002, Dunne et al. 2004), sta- estimates of grazing on seagrasses could reflect ‘run- tic measures of biomass within trophic compartments, away’ consumption by herbivores in locations where or mass balance approximations of the nutritional threats of attack by piscivores have been reduced to requirements of higher-order consumers to estimate ecologically trivial levels (cf. Mumby et al. 2006). Evi- production at the base of reef food webs. Given the dence supporting this second alternative comes from a permeable boundary between seagrasses and coral number of coral reefs where consumer reductions have reefs, it is likely that the importance of allochthonous been shown to have had dramatic cascading effects transfers of seagrass production to coral reefs has been on lower trophic levels (McClanahan et al. 1995, Mc- greatly underestimated. Clanahan 1998). If the second alternative is true, then In fact, empirical evidence supports this contention. the few existing studies of grazing on seagrasses are While many herbivorous fishes seek shelter on reefs at likely to have greatly overestimated historical levels of night, they commonly forage in nearby seagrass habitats consumption. throughout the day (Randall 1965, Ogden & Zieman The Florida Keys are intensively fished year-round, 1977, McAfee & Morgan 1996). Moreover, the conspicu- with seasonal peaks during the winter tourist season and ous presence of bare-sand ‘halos’ surrounding reefs cre- summer months (Bohnsack et al. 1994). During the ated by herbivores shows that grazing on nearby sea- 1970s, trap fishing was heavy throughout south Florida grasses can be intense (e.g. Randall 1965, Ogden & and concerns about wholesale reductions of fish stocks Zieman 1977, McAfee & Morgan 1996). Although the in- led to a 1980 ban of their use within 3 nautical miles off- tensity of grazing diminishes with increasing distance shore. Subsequently, trap fishing was also banned in from the reef (MacIntyre et al. 1987), grazing by reef- federal waters in depths of less than 30 m (Bohnsack et associated herbivores and seagrass-resident herbivores al. 1989). Because indiscriminant trap fishing has been is not limited to the areas immediately surrounding reefs banned in the Florida Keys, reef fishing has mostly (MacIntyre et al. 1987, Kirsch et al. 2002). Further, sea- targeted large piscivores, including amberjacks Seriola grass-resident herbivores are prey for a number of reef- dumerili, groupers (Serranidae), and to a lesser extent associated piscivores, increasing the transfer of seagrass barracudas Sphyraena barracuda, and some omnivores, production throughout coral reef food webs (Clifton & including hogfishes (Labridae), snappers (Lutjanidae), Robertson 1993, Kirsch et al. 2002). triggerfishes (Balistidae), and some grunt species There is now substantial anecdotal, descriptive, iso- (Haemulidae) (Bohnsack et al. 1994, Ault et al. 1998). topic, and experimental evidence that herbivores con- Herbivores are not reported to be targeted on reefs sume significant quantities of seagrass biomass in anywhere within the Florida Keys (Ault et al. 1998). many areas (see review by Valentine & Duffy 2006). In 1997, the US Congress designated an almost Even so, there is evidence that modern day grazing on 10 000 km2 area surrounding the Florida Keys as the seagrasses is anomalously low in subtropical and trop- Florida Keys National Marine Sanctuary (FKNMS) to ical ecosystems due to the recent overharvesting of protect its biodiversity and achieve sustainable use of green turtles and sirenians (summarized in Valentine reef resources. As a result, 23 replicate ‘no-take’ zones & Duffy 2006). While direct estimates of seagrass con- (designated as either ‘sanctuary preservation areas’ sumption by herbivores are few, a recent study has [SPAs] or, in 1 case, ‘ecological reserve’ [ER]) were es- shown that substantial amounts of seagrass production tablished along the Florida Keys reef tract (see Fig. 1). are consumed by smaller herbivores in the vicinity of The designation of these no-take zones limits the losses coral reefs (Kirsch et al. 2002). of most fishes to harvests at non-protected reefs. Valentine et al.: Overfishing impacts on seagrass–coral reef interface 39

At the time of this study, reviews (Halpern & Warner MATERIALS AND METHODS 2002, Halpern 2003, but see McClanahan & Graham 2005 and Russ & Alcala 2004) strongly indicated that Experimental design. To quantify the amount of the establishment of such no-take zones led rapidly (in seagrass production reaching coral reef food webs 1 to 3 yr) to substantial increases in the density, body via grazing and the extent to which the intensity of size, and biomass of most targeted fish species. A this trophic transfer is controlled by piscivores, we single study from the Florida Keys showed that reduc- used a simple experimental design that measured tions in fishing pressure led to an almost doubling of multiple response variables at 2 distances (immedi- snapper density and a 4-fold increase in the density of ately adjacent, [5 m] to the reef, which we term grunts only 2 yr after the establishment of Looe Key ‘near’ and at 30 or 60 m into the seagrass [Thalassia Reef as a no-take reef (Clark et al. 1989). In a second, testudinum] bed, which we term ‘far’) from repli- as yet unpublished study (Fioravanti & Valentine cated fished (predator-poor) and no-take (predator- 2007), far greater abundances of snappers were found rich) reefs. These reefs were located in 2 widely sep- at more recently established no-take sites in the upper arated regions of the Florida Keys reef tract (Fig. 1). Florida Keys (Dry Rocks and Grecian Rocks SPAs) than The present study was begun some 3 yr after the at a nearby fished reef (Little Grecian Reef). Increased establishment of no-take reefs within the Florida density of targeted fishes within such no-take zones Keys. As such, there should have been adequate has also been positively correlated with more abun- time for densities of reef piscivores to have recovered dant, and larger sized, targeted fishes in nearby fished prior to commencement of the study (Halpern & areas (reflecting the spillover effect) (McClanahan & Warner 2002, Halpern 2003). Mangi 2000, Roberts et al. 2001, but see Shipp 2003). As such, persistent, often intense, fishing on unprotected reefs of the FKNMS represents a sus- tained predator-reduction treatment, and the no-take zones (where the harvest of marine organisms is prohibited) a predator-rich treatment. Thus, establishment of no-take zones in the Florida Keys has provided a unique opportunity to test the extent to which the abundance of piscivorous fishes controls the flow of energy from seagrasses to coral reef food webs via the grazing pathway. Herein, we report the findings of a study that quantified grazing by both reef-associated and seagrass-resident herbivores in vegetated areas surrounding coral reefs in 2 widely separated regions of the Florida Keys. We also provide the first direct estimates of the quantity of seagrass production consumed by reef-associated parrotfishes and the extent to which the trophic exchange varies among locations and between regions. We also used the presence of areas protected from fishing to assess the extent to which the removal of piscivores from nearshore waters has altered the intensity of grazing across the seagrass–coral reef interface. Finally, we provide the first quantitative evaluation of the extent to which variation in herbivore guild Fig. 1. Study locations in upper and lower Florida Keys (20 m isobar shown). composition varies with location and pis- (q) Approximate locations of replicate study sites (see ‘Materials and methods’ civore density. for geographical coordinates). ER: Ecological reserve 40 Mar Ecol Prog Ser 333: 37–50, 2007

We predicted that there would be elevated densities Little Grecian Reefs, 25° 06.91’ N, 080° 18.20’ W; New- of piscivorous fishes foraging in seagrasses near pro- found Harbor, 24° 36.925’ N, 081° 23.642’ W; Looe Key, tected (no-take) coral reefs, and that this would lead to 24° 32.830’ N, 081° 23.642’ W; ER, 24° 28.879’ N, 081° reductions in density and shifts in the relative abun- 43.172’ W; patch site near ER, 24° 31.698’ N, 081° dances of grazers (from more larger reef-associated 39.205’ W; 2 sites along West Washerwoman Reef tract, parrotfishes to fewer more secretive, smaller, seagrass- 24° 32.890’ N, 081° 35.414’ W and 24° 33.316’ N, 081° resident parrotfishes) in the seagrass (Thalassia tes- 33.835’ W. tudinum), and that grazing on seagrass would be lower Response variables. Fish density and composition: near no-take reefs (Fig. 2). Conversely, we predicted Two commonly used visual census techniques were that lower densities of piscivorous fishes in fished used to document treatment effects on piscivore and areas would allow reef-associated herbivores to forage herbivore density and composition: point counts and freely throughout the seagrasses adjacent to fished belt transects. Visual census methods have been used reefs and that grazing intensity would be greater there extensively to describe the community structure and (Fig. 2). estimate the density of reef fishes (Bohnsack & Ban- To test these predictions, a navigation line was an- nerot 1986, Edgar & Barrett 1999). Although observer chored at a haphazardly located point at the base of each error and behavioral interactions between divers and replicate coral reef (see next subsection for specific sites fishes can bias the results of visual censuses (e.g. Sale & used in this study) and extended to distances of either Douglas 1981, Kulbicki 1998), direct counts are consid- 60 or 30 m into the adjacent seagrass bed in the upper ered to be reliable for most readily observable species. and lower Florida Keys, respectively; transect length was shortened in the lower Keys based on results from the first 2 yr of study in the upper Keys. The transect was established, with the aid of observers on the boat, such that it lay perpendicular to the perimeter of the reef. Data were collected in areas directly adjacent to the reef (hereafter referred to as the near site) and at the end of the transect (hereafter referred to as the far site). Data were collected in the upper Florida Keys in July and September 2000 and May and September 2001, and in the lower Florida Keys in May, July, and August 2002 and May and August 2003. Study sites. As far as possible, struc- turally similar fished and no-take reefs, with extensive seagrass habitats adjacent to the landward sides of the reef were selected for study. In the upper Keys, experiments were conducted in 2 no-take zones, Grecian Rocks and Dry Rocks SPAs, and at 2 fished sites, White Banks and Little Grecian Reefs (Fig. 1). In the lower Keys, experiments were conducted in the back reef environments of 3 no-take sites, Newfound Harbor, Looe Key SPAs and Western Sambos Ecological Reserve (ER) and 3 fished sites, 1 near the ER and 2 along the Washerwoman Reef Tract (Fig. 1). The geographical coordinates were Grecian Fig. 2. Predicted impacts at no-take sites of elevated piscivore reef fish density on seagrass (Thalassia testudinum) herbivores and grazing intensity in the Rocks, 25° 06.91’ N, 080° 18.20’ W; Dry sanctuary preservation areas. Diagram: scanned leaf image showing bite Rocks, 25° 07.59’ N, 080° 17.91’ W; White marks. Snapper, grouper and parrotfish images courtesy of R. Froese and Banks, 25° 02.442’ N, 080° 22.276’ W; D. Pauly (editors, FishBase 2006, www.fishbase.org, version 02/2006) Valentine et al.: Overfishing impacts on seagrass–coral reef interface 41

SCUBA-equipped observers made counts of pisci- in the upper, and Little Palm Island (24° 37.11’ N, vore- and reef-associated herbivore composition and 081° 24.42’ W), in the lower Keys. Leaves were severed density, at both the near and far sites, using a slightly from the collected shoots at the top of the sheath and modified version of the stationary point method of examined for bite marks. If bite marks were present, Bohnsack & Bannerot (1986). The only modification the shoots were severed below the bite mark to provi- was that 2 intersecting 10 m long lines were used to de uniformity in leaf appearance. Leaves were digital- mark the center point of an imaginary cylinder extend- ly scanned (Hewlett Packard®, flat-bed scanner), then ing upward from that point to the overlying water sur- rebundled into shoots and attached to a 0.25 m long face. During each sampling event, the diver knelt at sisal line with a clothespin. Each sisal line (or tether) the intersection of the 2 lines and identified (to the contained 3 evenly spaced shoots. Three lines of species level in most cases) and enumerated all fishes tethers, containing a total of 9 shoots, were placed at entering the area of the circle over a 15 min period. the specified distances from the reef at each study site. Counts were limited to days when visibility exceeded Tethers were anchored using wire stakes that were 20 m, and the divers conducting surveys at the near pushed into the sediment such that they simulated the and far sites were the same throughout each visit. appearance of natural shoots within a seagrass bed. Because piscivore and herbivore composition and den- Tethers were retrieved after 24 h and replaced with a sity varies greatly from day to day (Sale & Douglas new set of shoots. The remaining leaves on the re- 1981), visual surveys were repeated on 3 separate days trieved tethers were rescanned in the laboratory to at each distance and site during each of the previously estimate losses of leaf tissue due to grazing. This pro- reported survey times. cess was replicated for 3 d during each visit to the Belt transect surveys were conducted a few minutes study sites. after the point counts to provide a comparative esti- Grazing by large reef-associated herbivores—bite mate of the densities and compositions of herbivorous size: Because our point counts and belt transects were fishes. To do this, 20 × 1 m transect lines were estab- limited to a very small proportion of each survey day, lished parallel to the reef edge, at both near and far we evaluated the longer-term impacts of piscivore distances from the reef. Separate SCUBA-equipped density on grazing by both larger, reef-associated, par- observers swam the transect lines for a 5 min period rotfishes, and smaller, seagrass-resident parrotfishes with a 1 m long T-bar. All herbivores observed within by bite size. We measured bite size on randomly the 20 m2 area (as determined by the T-bar) were selected shoots from tethers and on shoots from pro- counted and identified. These counts were also reduction quadrats (see next subsection), both near and peated on 3 separate days at the near and far locations far from each of the reefs during each survey period. for the months, years, and sites described above (sub- Bite width, an indicator of herbivory by either a reef- section ‘Experimental design’). Piscivores were not associated or a seagrass-resident herbivore, was de- recorded in these surveys as we anticipated that belt termined from calibrated, scanned, leaf images that transect surveys would not accurately depict the densi- were imported into the commercially available Sigma ties of these highly vagile consumers. ScanTM digitizing software (cf. Kirsch et al. 2002). Once Grazing intensity: We quantified seagrass grazing scanned, bite widths were categorized as being due to intensity following methods described in Kirsch et al. feeding by either a larger reef-associated herbivore or (2002). To estimate the risk that reef-associated herbi- a smaller seagrass-resident herbivore. To provide a vores were willing to take in foraging in seagrass habi- benchmark for this categorization, undamaged turtle- tats adjacent to the reefs, tethered seagrass shoots (see grass leaves were placed in 37.8 l aquaria with a next paragraph) were placed at each site at distances known, abundant, seagrass-resident herbivore (adult of 5, 10, 20, 40, and 60 m from each reef in the upper bucktooth parrotfish Sparisoma radians) (Randall 1967). Keys, and at 5, 10, 20, 30 m from each reef in the lower The turtlegrass leaves were removed after 2 h and the Keys. Transect length was shortened in the lower Keys widths of the bites were determined as described as we found no evidence of statistically significant dif- above. Bite sizes for the bucktooth parrotfish ranged ferences in leaf loss at distances of 20, 40 and 60 m from 0.32 to 0.59 cm (n = 31). With the benchmark from reefs surveyed in the upper Keys. Tethers placed established, all bites with widths >0.59 cm were con- at distances of 5 and 10 m from the reef were grouped sidered to have been made by a larger reef-associated together to estimate grazing near the reef and tethers parrotfish, while smaller bite widths were assigned to placed at distances of 20, 30, 40, or 60 m were grouped the seagrass-resident category. together to estimate grazing away from the reef (cf. Seagrass density, production and aboveground bio- MacIntyre et al. 1987). mass: To quantify the amount of seagrass production Turtlegrass shoots Thalassia testudinum were col- consumed by herbivores, estimates of net above- lected from Rodriquez Key (25° 03.200’ N, 080° 27.1’ W) ground primary production (NAPP) and aboveground 42 Mar Ecol Prog Ser 333: 37–50, 2007

seagrass biomass were made from 3 replicate 0.01 m2 the 3 repeated measures of piscivore density, herbi- quadrats placed at distances of 5 m and either 30 m vore density, proportional losses of tethered leaves, (lower Keys) or 60 m (upper Keys) from the reef at each and the proportion of herbivore bites on tethered site and during each sampling period. leaves whose diameter was >0.59 cm. We estimated NAPP using a modified punch-hole Treatment effects on piscivore and herbivore den- technique (Kirkman & Reid 1979). Initially, turtlegrass sity and guild composition: Piscivores included in shoots within the haphazardly placed quadrats were comparisons were limited to those predators known to marked at the top of the leaf sheath; after 7 to 12 d, all eat seagrass herbivores (as documented by Randall marked shoots were removed, placed in prelabeled 1967 and McAfee & Morgan 1996). Those encountered plastic bags, and returned to the laboratory for pro- during our point count surveys were barracudas Sphy- cessing. In the laboratory, both the area and the bio- raena barracuda, rock hinds Epinephelus adscen- mass of the new and old growth, along with shoot den- sionis, black groupers Mycteroperca bonaci, mutton sity, were measured. The new growth was defined as snappers Lutjanus analis, yellow jacks Caranx barth- the area and weight of leaf tissue between the initial olomaei and bar jacks C. ruber (Table 1). Known sea- punch-hole scar and the top of the leaf sheath, plus any grass herbivores were limited to species of parrotfishes new unmarked leaves. The new growth and remaining in the genera Sparisoma and Scarus (Table 1). old growth were first scanned to determine the new Evaluations of treatment effects and their interac- leaf area produced and the total leaf area, and then all tions on piscivore and herbivore density were made the material was dried at 60oC to a constant mass. This using 3-way analyses of variance (ANOVAs), testing mass was standardized to estimate NAPP as g dry mass for differences between regions, between distances m–2 d–1. Total aboveground seagrass biomass was also from the reef, and between fished and no-take reefs. determined for these same quadrats. Data from point and belt transect surveys were either

Seagrass consumption by reef herbivores: To esti- square root- or log10-transformed data, respectively. mate cross-habitat energy exchanges at our study sites When no significant interactions were detected, they and the extent to which trophic transfer varied both were removed from the model and the analyses were within and between regions, we first estimated reef repeated. If significant interactions were detected, perimeter from ‘Geographic Information System’ GIS- comparisons of main effects impacts on the dependent based maps of each reef using ARCVIEW software. variable of interest are simply discussed. This estimate was then multiplied by 10 (the farthest In some cases, violations of homogeneity of variance distance in meters at which large parrotfish grazing were the direct result of the high variance to mean was detected), to provide a conservative estimate of ratios for episodically encountered schooling fishes. the area over which reef-associated herbivores were Since there is no nonparametric analog of 3-way feeding. Next, we estimated the quantity of seagrass production transferred to the reef via grazing by con- Table 1. Piscivores observed in point and belt counts and verting site-specific daily rates of areal leaf loss on our known to be predators of parrotfish (Randall 1965, 1967, tethers (calculated on a per shoot basis) to biomass on McAfee & Morgan 1996), and seagrass herbivores (parrotfishes) observed in point and belt counts a per shoot basis (y = 0.2228 + 190.08x, where y = bio- 2 mass lost per shoot and x = leaf loss in cm ; F156 = 931.145, p < 0.0001, r2 = 0.86). We then multiplied this Common name Scientific name value by the estimated total number of shoots along Piscivores the perimeter of the reef, which was based on mean Barracuda Sphyraena barracuda shoot density determined from the quadrats. From the Rock hind Epinephelus adscensionis NAPP estimates, daily estimates of the area of new leaf Black grouper Mycteroperca bonaci growth (cm2) shoot–1 were regressed on the estimate of Mutton snapper Lutjanus analis Yellow jack Caranx bartholomaei leaf mass (g). These daily estimates of production were Bar jack Caranx ruber then converted to an annual estimate of cross-habitat energy exchange by multiplying by 365 d. Seagrass herbivores Statistical analyses. Replicate definition: Main treat- Redband Sparisoma aurofrenatum Stoplight Sparisoma viride ment effects included region (sites in the upper and Yellowtail Sparisoma rubripinne lower Florida Keys), fishing pressure (fished and no- Rainbow Scarus guacamaia take sites) and distance from the reef (near and far). To Bucktooth Sparisoma radians evaluate our predictions, we repeated data collection 2 Midnight Scarus coelestinus to 3 times in each of the 2 yr study periods (see sub- Redtail Sparisoma chrysopterum Unidentified small section ‘Experimental design’) at all sites. Estimates of parrotfishes different variables were generated from the mean of Valentine et al.: Overfishing impacts on seagrass–coral reef interface 43

ANOVA, the results of 3-way ANOVA and the viola- Synergistic relationships: We used both Kendall’s tion of the assumption are reported. In cases where the tau-b and Pearson’s product moment correlation ana- null hypothesis was not rejected, the analyses are sim- lyses on untransformed data to evaluate the predic- ply presented since ANOVA is robust to violations of tions regarding the impacts of piscivore removal on homogeneity of variance with respect to Type II errors seagrass herbivory (herbivore density, grazing, sea- (failing to reject a false null hypothesis; Underwood grass growth, biomass, and density). When significant 1981). In cases where the main effects or their inter- (p < 0.05) positive correlations between variables of actions were significant, treatment impacts were com- interest were detected, the findings are reported. pared by plotting the data. Comparisons of main effect When negative relationships were detected, the rela- impacts on the composition of the piscivore and her- tionship between the transformed (either square root- bivore guilds were made using the nonparametric or log10-transformed) variables was evaluated further analysis of similarity technique (ANOSIM; 4th root using simple linear regression. transformed, Bray-Curtis similarity technique) (Clarke 1993). Results in all comparisons were considered to be sig- RESULTS nificant when p < 0.05. All parametric analyses were conducted using SPSS™ Version 11.0 statistical soft- Treatment effects on piscivore density ware package. ANOSIM and SIMPER comparisons were made using the multivariate nonparametric soft- Three piscivore species (bar jacks Caranx ruber, ware package Primer-E™. yellow jacks C. bartholomaei, and great barracuda Treatment effects on grazing intensity and bite size: Sphyraena barracuda) accounted for ~98% of all Each tether was considered to be a replicate and the individuals recorded during point count surveys. average amount of the tissue lost, on a per shoot basis, Three others (rock hinds Epinephelus adscensionis, on the 3 tethers provided a daily estimate of grazing at black groupers Mycteroperca bonaci, and mutton each site and distance during a survey. Daily areal snappers Lutjanus analis) were encountered only losses of tissue were calculated by multiplying this loss at no-take reefs and their total abundance was by the average density of shoots estimated from the <10 individuals. Therefore, evaluations of treatment NAPP quadrats (cf. Kirsch et al. 2002). The areal losses effects were limited to the 3 most abundant pisci- were then divided by the areal estimate of leaf produc- vores. tion to provide an estimate of grazing intensity. The Bar jacks were by far the most abundant of the pisci- results of this conversion provided a standardized esti- vores, and we encountered them mostly in dense mate of grazing intensity that can easily be compared schools (variance to mean ratios ranging from 74:1 to with previously published estimates of the proportion 15:1 in the upper and lower Florida Keys, respectively). of seagrass production consumed by herbivores Bar jacks were far more abundant at near than at far

(Valentine & Duffy 2006). Main treatment effects sites (log-transformed counts: F1,68 = 7.64; p < 0.007; impacts (region, fishing pressure, and distance from Fig. 3a). There were no other significant treatments the reef) on grazing intensity were tested using a effects on bar jack density, nor were there significant 3-way ANOVA on arcsine square root transformed interactions between treatments. estimates of the proportion of NAPP consumed by Great barracuda were encountered in smaller grazers. Similarly, the impacts of the main effects on schools than bar jacks (variance to mean ratios of the daily rates of grazing by larger reef-associated 5:1 to 9:1) (Fig. 3b). This resulted in the point count grazers on seagrasses adjacent to the reef were tested data violating the ANOVA homogeneity of vari- using a 3-way ANOVA of arcsine square root transfor- ance assumption, regardless of transformation. Log- mations of the proportions of tethered shoots with bite transformed counts of great barracuda were higher widths >0.59 cm. in the lower than upper Keys (F1,68 = 4.62, p < 0.035; Treatment effects on NAPP and aboveground bio- Fig. 3b), and more so near than far from the reefs mass: Comparisons of NAPP and aboveground sea- (F1,68 = 6.49; p < 0.013; Fig. 3b). No other significant grass biomass were analyzed using 3-way ANOVA, effects on great barracuda density were detected with interactions, region (upper vs. lower Keys), fish- (p > 0.05). ing pressure (fished vs. no-take), and distance (near Yellow jacks were encountered infrequently (<5% vs. far) as factors. If the data did not satisfy the of all counts), and in much lower densities than the assumption of homogeneity of variance for ANOVA, 2 other species of piscivores. While yellow jacks main effects comparisons for NAPP and aboveground were more abundant far from reefs in the upper biomass were made following log transformation of Keys, the opposite pattern was true in the lower Keys the data. (Fig. 3c). 44 Mar Ecol Prog Ser 333: 37–50, 2007

Bar jacks a Fishing pressure had a significant effect on square 40 * Near root transformed counts of seagrass-resident parrot- 30 Far fishes (F1,62 = 12.97; 62; p < 0.001; Fig. 4b): These fishes were about twice as abundant at no-take reefs than at 20 * fished reefs. No other significant main effects or interactions were detected between regions or between 10 * * distances from the reef. Herbivore composition varied significantly between 0 regions (2-way crossed ANOSIM: global R = 0.097; p = 0.023) and with levels of fishing pressure (global R = 8 Great barracuda b 0.154; p = 0.04). More reef-associated species of parrot- A * fishes were recorded in point counts in the lower Keys. 6 * SIMPER analysis detected modest differences in * redtail and stoplight parrotfish abundances between 4 B regions. Additionally, counts of rainbow and yellowtail * parrotfishes were limited to sites in the lower Keys. 2 A second 2-way crossed ANOSIM showed that small Number counted parrotfish (mostly bucktooth) species composition also 0 varied significantly between regions (global R = 0.132; p = 0.001) and with levels of fishing pressure (global 4 Yellow jacks c

Large parrotfishes a 3 50 * Near 2 Far 40

1 30 0 FishedNo-take Fished No-take 20 * Upper Keys Lower Keys * * 10 Fig. 3. Distributional pattern of dominant piscivores during point counts in upper and lower Florida Keys; ‘Near’ = 5 m and ‘Far’ = 60 m (upper Keys) or 30 m (lower Keys) into sea- 0 grass bed. Uppercase letters denote statistically significant b differences between regions, asterisks statistically significant Small parrotfishes 140 differences between distances from reef. Values are treatment means + 1 SE

Number counted 120

100 Treatment effects on herbivore composition and density 80 60 Parrotfishes recorded during our surveys fell into 2 categories: schools of larger reef-associated (yellow- 40 fin, redtail, redfin, princess, stoplight, midnight and 20 rainbow), and smaller seagrass-resident parrotfishes (bucktooth and unidentified juvenile parrotfishes). 0 Fished No-take Fished No-take While log10-transformed counts of large parrotfishes failed to satisfy the assumption of homogeneity of vari- Upper Keys Lower Keys ance for ANOVA, significant differences between dis- Fig. 4. Distributional patterns of (a) larger reef-associated and tances existed (F1,61 = 24.89; p < 0.0001). Overall there (b) smaller seagrass-resident parrotfishes during point counts were more large parrotfishes in the no-take zones, but at study sites in upper and lower Florida Keys. All differences between fished and no-take reefs were significantly different this was only apparent in the lower Keys (interaction at p < 0.05; (*) Statistically significant differences between between region and fishing pressure: F1,61 = 15.91; p < distances from reef; for definition of ‘Near’, ‘Far’ see Fig. 3. 0.001) (Fig. 4a). Values are treatment means + 1 SE Valentine et al.: Overfishing impacts on seagrass–coral reef interface 45

R = 0.105; p = 0.005). In addition, the small parrotfish composition of large parrotfishes did not vary with species composition differed significantly with dis- fishing pressure (global R = 0.013; p = 0.292) or region tance from the reef (global R = 0.08; p = 0.015). (global R = 0.02; p = 0.222). While differences in herbi- Stoplight, rainbow, and yellowtail parrotfishes ac- vore composition between distances were apparent, counted for 71.4, 19.4 and 9.2% of grazers recorded, they were not significant (global R = 0.02; p = 0.056). respectively, in the belt transect surveys. Redtail Log-transformed counts of seagrass-resident parrot- parrotfish were not recorded on the belt transects. fishes varied significantly with fishing pressure (F1,77 = Large parrotfish were encountered in schools. Vari- 14.95; p < 0.0001). Densities of these smaller parrotfishes ance to mean ratios ranged from 14.5:1 and from 21:1 were on average 5 times greater at no-take sites, regard- for yellowtail and stoplight parrotfishes, respectively. less of distance from the reef, than at the fished sites As a result, the assumption of homogeneity of variance (2.8 parrotfish vs. 0.42 per 20 m2), and markedly so in the for ANOVA was violated. upper Keys (Fig. 5b). No significant differences were Log-transformed counts of reef-associated parrot- detected between regions or distances from the reef. fishes on the belt transects diminished significantly Small parrotfish composition varied significantly with with increasing distance from the reef (F1,77 = 6.45; p < region (global R = 0.115, p = 0.017) and fishing pressure 0.02), being about twice as abundant at near than at far (global R = 0.168, p = 0.001). As a result, additional sites (Fig. 5a). Large reef-associated parrotfishes were crossed (treatment × distance) ANOSIMs were con- also significantly more abundant at no-take than at ducted on surveys from the upper and lower Keys. fished sites (F1,77 = 5.28; p < 0.024), regardless of region Compositions of these parrotfishes varied greatly with (Fig. 5a). Stoplight and rainbow parrotfishes were most fishing pressure in both regions (lower Keys: global R = abundant near reefs. The results of ANOSIM differed 0.245; p = 0.001, upper Keys: global R = 0.227, p = somewhat from those of the point surveys in that the 0.024). These differences do not seem to be ecologically relevant, as they are best explained by the higher pro- Large parrotfishes a portion of unidentifiable small parrotfishes observed at 16 * the no-take sites. At the fished sites in the lower Keys, 14 Near there were 292 bucktooths and only 17 unidentified Far parrotfishes, whereas in the no-take areas we observed 12 979 bucktooths and 230 unidentified individuals. 10 Comparisons of herbivores recorded in the 2 survey 8 types produced complementary results. There was a high degree of overlap in the species composition of 6 parrotfishes recorded by the 2 approaches. All but red- 4 * tail and midnight parrotfishes were recorded in both 2 * * surveys. Of these, only redtail parrotfish were recorded in abundance, so it seems likely that we ade- 0 quately characterized the herbivore guild at the study 35 Small parrotfishes b sites. Significant positive correlations between counts 30 made using point and belt transect survey methods Number counted were found for stoplight (tau-b = 0.498, p < 0.001), 25 yellowtail (tau-b = 0.463, p < 0.007), and rainbow 20 (tau-b = 0.396, p < 0.018) parrotfishes, indicating that our characterizations of treatment effects on the grazer 15 density and composition were consistent. 10

5 Treatment effects on grazing intensity

0 Fished No-take Fished No-take Grazing intensity varied significantly with region (F1,68 = 21.13; p < 0.0001; Fig. 6a), and distance from the Upper Keys Lower Keys reefs (F1,77 = 4.59; p < 0.036) and was higher in the up- Fig. 5. Distributional patterns of (a) larger reef-associated and per than in the lower Keys, and higher near than far (b) smaller seagrass-resident parrotfishes during belt counts from reefs. Grazing was over 200% > seagrass produc- at study sites in upper and lower Florida Keys. All differences between fished and no-take reefs were significantly different tion near reefs in the upper Keys (Fig. 6a). In stark con- at p < 0.05. Symbols and definitions as in Figs. 3 & 4. Values trast, grazing never exceeded 48% of seagrass produc- are treatment means + 1 SE tion in the lower Keys. Grazing was also substantial at 46 Mar Ecol Prog Ser 333: 37–50, 2007

–2 400 Near a Nor was the correlation between proportions of bites

m * Far –1 by large parrotfishes on tethers and on shoots from the A production quadrats (tau-b = –0.118, p > 0.43). * Using GIS-based estimates of reef perimeter, bite widths on tethers, and daily rates of leaf loss on 300 tethered shoots, we estimated that, on average, 9.2 t of 100 seagrass production (range 0 to 18.28) are eaten by larger reef-resident parrotfishes each year at the reefs examined in the upper Keys. For the lower Keys, the B * estimates were much lower, averaging around 0.75 t * yr–1 (range: 0 to 2.422) at the reefs examined in this 0 Percent NAPP consumed d study.

120 –1 b Treatment effects on NAPP and aboveground 100 * * biomass * 80 * Log-transformed estimates of daily rates of produc- 60 tion showed that NAPP was significantly greater in the

lower than in the upper Keys (F1,74 = 11.97; p < 0.001; 40 Fig. 7a), but did not vary significantly with distance from reefs or level of fishing pressure. Log-trans- 20 formed estimates of aboveground biomass were signif-

Percent of tethers attacked d 0 Fished No take Fished No take 0.06 a Upper Keys Lower Keys Near A Far Fig. 6. Fishing effects on proportion of (a) net aboveground ) 0.05 primary production (NAPP) consumed and (b) seagrass teth- –2 ers attacked at varying distances from fished and unfished 0.04 reefs in upper and lower Florida Keys. Near: 5 to 10 m; Far: 20 B and 30 m (lower Keys) or 20, 40, 60 m (Upper Keys) Symbols 0.03 as in Fig. 3. Values are treatment means + 1 SE 0.02 far sites (ranging between one half and two-thirds of 0.01 the measured seagrass production) in the upper Keys. Production (g dw m 0.00 0.16 Treatment effects on parrotfish bite size * 0.14 b ) * –2 * The proportions of tethered shoots and shoots har- 0.12 vested from quadrats that were bitten by larger reef- * associated herbivores were similarly high near reefs 0.10 (>60 and >80% of shoots on tethers or harvested from 0.08 quadrats, respectively) (Fig. 6b). The number of bites 0.06 by large herbivores varied significantly only with dis- 0.04 tance from the reefs (F1,62 = 24.53; p < 0.0001). (g dw m Biomass Regional differences in grazing were due to differ- 0.02 ences in feeding by reef-associated parrotfishes. The 0.00 correlation between the arcsine-transformed propor- FishedNo take Fished No take tions of leaf losses on tethers and proportions of teth- Upper Keys Lower Keys ered shoots bitten by large parrotfishes was significant (tau-b = 0.431; p < 0.001), but the correlation between Fig. 7. Thalassia testudinum. Net aboveground (a) primary production and (b) biomass of seagrass near and far from the arcsine-transformed proportions of tissue loss on reefs. For definition of ‘near’ and ‘far’ see Fig. 6. Symbols and tethers and numbers of bites on shoots harvested from definitions as in Fig. 3; dw = dry weight. Values are treatment production quadrats was not (tau-b = 0.05, p > 0.71). means + 1 SE Valentine et al.: Overfishing impacts on seagrass–coral reef interface 47

icantly lower near than far from reefs (F1,74 = 10.59; p < sites) are eaten by larger reef-resident parrot fishes 0.002; Fig. 7b). No other significant main effects or each year at the study sites in the upper Keys. The interactions were detected. magnitude of this transfer can apparently vary greatly among regions as these estimates are much lower, averaging around 0.75 t yr–1 (range 0 to 2.422) for the Piscivore impact on herbivore density and seagrass lower Keys. Given the high levels of grazing by smaller herbivores, which are known prey for a number of Because significant treatment effects were primarily reef-associated piscivores in the upper Keys, it is likely limited to areas near reefs, and because of the absence that our estimates of energy transfer from seagrasses to of the anticipated treatment effect on piscivore density, reefs via grazing are low in the upper Keys. we refocused our attention on data collected from sites Marine herbivores have never been observed to near reefs. graze uniformly across any single habitat or landscape The anticipated significant negative correlations (e.g. Ogden et al. 1973, Hay 1981) and this was cer- between log-transformed densities of barracuda and tainly not the case in our study. At the landscape scale, parrotfishes were not detected. Nor did we find the seagrass grazing was 5 times greater in the upper than anticipated significant positive correlations between in the lower Keys. At the scale of the reefs, attacks by log-transformed densities of any species of parrotfishes large herbivores on tethered shoots and on shoots har- and proportional losses of leaf tissue on the tethers. We vested from quadrats were similarly high near reefs did, however, find a significant positive correlation (>60 and >80% of tethered shoots and shoots har- between the arcsine-transformed proportions of teth- vested from quadrats, respectively). Grazing was also ered shoots with bites made by larger parrotfishes and substantially greater away from reefs in the upper the arcise-transformed proportions of leaf tissue lost Keys (ranging between one half and two-thirds of the from the tethers (r = 0.489, p < 0.004, n = 33). A nega- measured seagrass production). tive correlation between the arcise-transformed losses If modern day estimates of grazing on seagrasses are of leaf tissue on the tethers and log-transformed NAPP anomalously high due to the truncation of reef food (r = –0.321, p < 0.078, n = 33) was not significant, nor webs by fishers, we should have found dramatic differ- was a negative correlation between great barracuda ences in herbivory on seagrasses between sites pro- density and arcise-transformed losses of leaf tissue on tected from fishing and those that are not. This was not tethered shoots (r = –0.321, p < 0.068, n = 33). Addi- the case: rates of herbivory on seagrasses were not tional regressions suggested that linear relationships much lower at sites protected from fishing than at sites between visual estimates of piscivore density, the pro- where fishing continues. While the composition of portion of shoots attacked by larger reef-associated parrotfishes varied significantly with distance from parrotfishes, leaf loss on tethers, and seagrass growth reefs, there was no evidence that differences in density were not strong (p > 0.05). or composition of reef-associated herbivores between regions contributed to regional differences in grazing. The significant regional differences in herbivore com- DISCUSSION position detected in ANOSIM analyses were simply the result of shifts in the relative abundances of the Despite emerging evidence showing the importance 2 numerically dominant species recorded in the point of exogenous inputs of nutrients to coral reef food webs counts rather than large differences in species com- (e.g. Leichter et al. 2003), most investigators adhere to position. Unexpectedly, large reef-associated parrot- the view that reef food webs are closed entities charac- fishes were significantly more abundant at no-take terized by tight recycling processes (Hatcher 1997). than at fished sites, regardless of region. This finding is Qualitatively, we know that trophic links exist be- similar to that of Mumby et al. (2006), who found ele- tween reefs and surrounding habitats, yet we know of vated densities of large parrotfishes in a predator-rich no direct estimates of the intensity of this exchange marine reserve in Bahamas. The mechanism proposed (reviewed in Bascompte et al. 2005, Valentine & Duffy to explain this abundance of grazers was protection 2006). Before the present study there were no known from collection by trap fisherman. This explanation direct estimates of how much seagrass production does not, however, explain the elevated abundances of reaches reef food webs via grazing in seagrass mead- large parrotfishes at no-take sites in the Florida Keys, ows. Using GIS-based estimates of reef perimeter at as trap fishing has been banned there since 1980 our study sites, measurements of bite widths on tethers (Bohnsack et al. 1989). and daily rates of tethered leaf loss, we estimated, on Still, we cannot discount the role that piscivores may average, that 9.2 t of seagrass production (ranging play in controlling grazing on seagrasses. Large differ- from 0 to 18.28 t of seagrass production among the 4 ences in grazing intensity between regions could have 48 Mar Ecol Prog Ser 333: 37–50, 2007

been the result of regional differences in piscivore (Colin 1996, Kulbicki 1998). Still, both point counts and abundance, especially of great barracuda, which were belt transect surveys have been successfully used to twice as abundant in the lower Keys. Although not document the benefits of establishing no-take zones prized as a commercial fish, great barracuda are elsewhere (Edgar & Barrett 1999). Thus, it seems that esteemed by some anglers as a game fish (Schmidt exploited species (both piscivores and omnivores) are 1989). As such, this species may be impacted by fishing truly more abundant at the no-take sites than at the along reefs in the northern Florida Keys. A negative fished sites. It is also possible that targeted piscivores significant correlation (r = –0.321) between great bar- were abundant in these seagrass habitats at night and racuda density and leaf losses on tethers, coupled with were missed during our surveys (Eggleston et al. 1998), the greater abundances of larger parrotfishes near although another study addressing a separate question reefs, suggests that predation risk (e.g. Heithaus et al. at our lower Keys study sites found smaller barjacks to 2002, Preisser et al. 2005) may limit large parrotfishes be the only piscivore present during nocturnal fish to feeding in seagrasses adjacent to reefs in the lower surveys (Blackmon 2005). Keys. Perhaps this explains why we did not detect a In conclusion, while the lack of effect of fishing pres- strong relationship between large parrotfish density sure on densities of large piscivores was unexpected, and the proportion of shoots attacked by them. Still, we found some evidence that the anticipated increases the modest correlation between tether losses and den- in regional densities of piscivorous fishes were associ- sities of great barracuda estimated from point counts ated with regional differences in grazing intensity by shows that if risk of predation plays a role in determin- parrotfishes. We also found that while grazing inten- ing grazing intensity on seagrasses surrounding the sity on seagrasses varied greatly, significant quantities reefs of the lower Keys, it is not a major one. A recent of seagrass production are reaching reef-resident con- evaluation of Caribbean reef food webs concluded that sumers. Taken together with emerging evidence on strong interactions between adjacent trophic levels are the importance of allochthonous inputs of nutrients for rare (Bascompte et al. 2005). As such, strong top–down algal growth, it seems clear that the existing paradigm restrictions on the transfer of seagrass production via of coral reef food webs as tightly coupled, recycling the grazing pathway are likely to be rare. networks requires re-examination. It is also clear that Based on reviews that documented higher densities more detailed quantification of cross-habitat energy of exploited piscivores in marine protected areas exchanges is needed to validate existing food web (Halpern & Warner 2002, Halpern 2003) and more spe- models. Among the important consequences of such cific regional studies that found higher densities of validation will be the ability to provide data-based targeted piscivores on protected reefs in the Florida landscape recommendations for establishing ‘marine Keys (e.g. Sluka et al. 1994), we anticipated recording protected area’ designs. greater abundances of these predators at the no-take sites than at fished sites. Instead, the piscivore guild Acknowledgements. The development of these ideas was was dominated by moderately fished, transient preda- supported by grants from University of North Carolina at tors (jacks and great barracuda) rather than heavily Wilmington National Undersea Research Center (UNCW No. fished snappers and groupers. Given that historical 9537), The Nature Conservancy’s Ecosystem Research Pro- data on these species are lacking (cf. Edgar & Barrett gram and Marine Fisheries Initiative Program (MARFIN). We also express our appreciation to Drs. J. Bohnsack, B. Causey, 1999), and the fact that persistent intense commercial B. Keller, and S. Miller, and UNCW Captains K. Boykin and fishing throughout the Keys pre-dated the early 1900s O. Rutten for providing us with a historical background on when investigations began (Viele 2001), it is uncertain fishing and the establishment of marine reserves in the if the conditions described in the present study at pro- Florida Keys National Marine Sanctuary. Additionally, we tected sites are representative of reef food webs in ear- thank Drs. R. Aronson, J. Cebriàn, S. Powers, and 4 anony- mous reviewers for constructive criticisms of this work. This lier times (Knowlton 2004). It is noteworthy, however, is MESC Contribution No. 368. that Hixon & Carr (1997) also found transient predators (bar jacks), and not snappers and groupers, to be the dominant piscivores at Lee Stocking Island, Bahamas, LITERATURE CITED where fishing pressure was reported to be low. Ault J, Bohnsack J, Meester G (1998) A retrospective (1979– It may also be that methodological artifacts (e.g. 1996) multispecies assessment of coral reef fish stocks in diver avoidance or short survey duration) led to low the Florida Keys. Fish Bull 96:395–414 incidences of targeted piscivores at no-take sites (Kul- Bascompte J, Melián CJ, Sala E (2005) Interaction strength bicki 1998, Edgar et al. 2004). Moreover, while some combinations and the overfishing of a marine food web. Proc Natl Acad Sci USA 102:5443–5447 piscivores tend to be widely dispersed, others actively Blackmon DC (2005) Diel vertical migration of seagrass-asso- seek shelter within the reef matrix and would not be ciated benthic invertebrates: a novel escape mechanism detected without the use of a roving diver technique that provides an allochthonous input of seagrass-based Valentine et al.: Overfishing impacts on seagrass–coral reef interface 49

production to coral reef resident predators, PhD disserta- Kirsch KD, Valentine JF, Heck KL (2002) Parrotfish grazing on tion, University of South Alabama, Mobile turtlegrass Thalassia testudinum: evidence for the impor- Bohnsack JA, Bannerot SP (1986) A stationary visual census tance of seagrass consumption in food web dynamics of technique for quantitatively assessing community struc- the Florida Keys National Marine Sanctuary. Mar Ecol ture of coral reef fishes. NOAA Tech Rep NMFS 41 Prog Ser 227:71–85 Bohnsack JA, Sutherland DL, Harper DE, McClellan DB, Knowlton N (2004) Multiple stable states and the conser- Hulsbeck MW, Holt CM (1989) The effects of fish trap size vation of marine ecosystems. Prog Oceanogr 60:387–396 on reef fish catch off southeastern Florida. Mar Fish Rev Kulbicki M (1998) How acquired behavior of commercial reef 51:36–46 fishes may influence the results obtained from visual Bohnsack JA, Harper DE, McClellan DB (1994) Fisheries trends censuses. J Exp Mar Biol Ecol 222:11–30 from Monroe County, Florida. Bull Mar Sci 54:982–1018 Leichter J, Stewart HL, Miller SL (2003) Episodic nutrient Cadenasso ML, Pisckett STA, Bell SS, Benning TL, Carreiro transport to Florida coral reefs. Limnol Oceanogr 48: MM, Dawson TE (2003) An interdisciplinary and syn- 1394–1407 thetic approach to ecological boundaries. BioScience 53: Link J (2002) Does food web theory work for marine eco- 717–722 systems? Mar Ecol Prog Ser 230:1–9 Clark JR, Causey B, Bohnsack JA (1989) Benefits from coral Macintyre IG, Graus RR, Reinthal PN, Littler MM, Littler DS reef protection: Looe Key Reef, Florida. In: Magoon (1987) The barrier reef sediment apron: Tobacco Reef, HCOT, Minor D, Tobin LT, Clark D (eds) Coastal zone ‘89 Belize. Coral Reefs 6:1–12 American Society of Civil Engineers, New York, Charles- McAfee ST, Morgan SG (1996) Resource use by five sym- ton, SC, p 3076–3086 patric parrotfishes in the San Blas Archipelago, Panama. Clarke K (1993) Non-parametric multivariate analyses of Mar Biol 125:427–437 changes in community structure. Aust J Ecol 18:117–143 McClanahan T (1998) Predation and the distribution and Clifton KE, Robertson DR (1993) Risks of alternative mating abundance of tropical sea urchin populations. J Exp Mar tactics. Nature 366:520 Biol Ecol 218:231–255 Colin PL (1996) Longevity of some coral reef fish spawning McClanahan T, Graham NAJ (2005) Recovery trajectories of aggregations. Copeia 1996:189–191 coral reef fish assemblages in Kenyan marine protected Dunne JA, Williams RJ, Martinez ND (2004) Network struc- areas. Mar Ecol Prog Ser 294:241–248 ture and robustness of marine food webs. Mar Ecol Prog McClanahan T, Mangi S (2000) Spillover of exploitable fishes Ser 273:291–302 from a marine park and its effect on the adjacent fishery. Edgar G, Barrett N (1999) Effects of the declaration of marine Ecol Appl 10:1792–1805 reserves of Tasmanian reef fishes, invertebrates, and McClanahan TR, Kamukuru AT, Muthiga NA, Gilagabher M, plants. J Exp Mar Biol Ecol 242:107–144 Obura D (1995) Effect of sea urchin reductions on algae, Edgar GJ, Bustamante RH, Fari JM, Calvopi M, Martinez NA, coral, and fish populations. Conserv Biol 10:136–154 Toral-Granda MV (2004) Bias in evaluating the effects of Mumby PJ, CP Dahlgren, AR Harborne, CV Kappel and 10 marine protected areas: the importance of baseline data others (2006) Fishing, trophic cascades, and the process of for the Galápagos Marine Reserve. Environ Conserv 31: grazing on coral reefs. Science 311:98–101 212–218 Myers R, Worm B (2003) Rapid worldwide depletion of pre- Eggleston DB, Grover JJ, Lipcius RN (1998) Ontogenetic diet datory fish communities. Nature 423:280–283 shifts in Nassau grouper: trophic linkages and predatory Nakano S, Murakami M (2001) Reciprocal subsidies: dynamic impacts. Bull Mar Sci 63:111–126 interdependence between terrestrial and aquatic food Fioravanti G, Valentine JF (2007) Impacts of reef architecture webs. Proc Natl Acad Sci USA 98:166–170 and fishing on food webs in back reef environments of Ogden JC, Zieman JC (1977) Ecological aspects of coral the upper Florida Keys national marine sanctuary. Fla Sci reef–seagrass bed contacts in the Caribbean. Proc 3rd (in press) Int Coral Reef Symp:377–382 Halpern BS (2003) The impact of marine reserves: do reserves Ogden JC, Brown R, Salesky N (1973) Grazing by the echi- work and does reserve size matter? Ecol Appl 13:117–137 noid Diadema antillarum: formation of halos around West Halpern BS, Warner RR (2002) Marine reserves have rapid Indian patch reefs. Science 182:715–717 and lasting effects. Ecol Lett 5:361–366 Pandolfi JM, Bradbury RH, Sala E, Hughes TP and 8 others Hatcher B (1997) Coral reef ecosystems: how much greater is (2003) Global trajectories of the long-term decline of coral the whole than the sum of the parts? Proc 8th Int Coral reef ecosystems. Science 301:955–958 Reef Symp 1:43–56 Polis G, Strong D (1996) Food web complexity and community Hay ME (1981) Spatial patterns of grazing intensity on a dynamics. Am Nat 147:813–846 Caribbean barrier reef: herbivory and algal distribution. Powers ME (2001) Prey exchange between a stream and its Aquat Bot 11:97–109 forested watershed elevates predator densities in both Heithaus MR, Dill LM, Marshall GJ, Buhleier B (2002) Habitat habitats. Proc Natl Acad Sci USA 98:14–15 use and foraging behavior of tiger sharks (Galeocerdo Preisser EL, Bolnick DI, Benard MF (2005) Scared to death? cuvier) in a seagrass ecosystem. Mar Biol 140:237–248 The effects of intimidation and consumption in predator– Hilderbrand GV, Hanley TA, Robbins CT, Schwartz CC prey interactions. Ecology 86:501–509 (1999) Role of brown bears (Ursus arctos) in the flow of Randall JE (1965) Grazing effect on sea grasses by herbi- marine nitrogen into a terrestrial ecosystem. Oecologia vorous reef fishes in the West Indies. Ecology 46:255–260 121:546–550 Randall JE (1967) Food habits of reef fishes of the West Indies. Hixon MA, Carr MH (1997) Synergistic predation, density Stud Trop Oceanogr 5:665–847 dependence, and population regulation in marine fish. Roberts C, Bohnsack JA, Gell F, Hawkins JP, Goodridge R Science 277:946–949 (2001) Effects of marine reserves on adjacent fisheries. Kirkman H, Reid D (1979) A study of the role of seagrass Posi- Science 294:1920–1923 donia australis in the carbon budget of an estuary. Aquat Russ GR, Alcala AC (2004) Marine reserves: long-term Bot 7:173–183 protection is required for full recovery of predatory fish 50 Mar Ecol Prog Ser 333: 37–50, 2007

populations. Oecologia 138:622–627 Underwood AJ (1981) Techniques and analysis of variance in Sale PF, Douglas W (1981) Precision and accuracy of visual experimental marine biology and ecology. Oceanogr Mar census technique for fish assemblages on coral patch Biol Annu Rev 19:513–605 reefs. Environ Biol Fish 6:333–339 Valentine JF, Duffy JE (2006) The central role of grazing in Schmidt TW (1989) Food habits, length–weight relationship seagrass ecology. In: Larkum AWD, Orth RJ, Duarte CM and condition factor of young great barracuda, Sphyraena (eds) Seagrass: biology, ecology and their conservation. barracuda (Walbaum), from Florida Bay, Everglades Kluwer Academic Publishers, Dordrecht p 463–501 National Park, Florida. Bull Mar Sci 44:163–170 Valentine JF, Heck KL (2005) Interaction strength at the coral Shipp RL (2003) A perspective on marine reserves as a fishery reef–seagrass interface: has overfishing diminished the management tool. Fish Res 28:12–21 importance of seagrass habitat production for coral reef Sluka R, Chiappone M, Sullivan KM (1994) Comparison of food webs? Coral Reefs. 24:209–213 juvenile grouper populations in southern Florida and the Viele J (2001) The Florida Keys. Vol 1. A history of the pio- central Bahamas. Bull Mar Sci 54:871–880 neers. Pineapple Press, Sarasota, FL

Editorial responsibility: Otto Kinne (Editor-in-Chief), Submitted: September 21, 2005; Accepted: July 3, 2006 Oldendorf/Luhe, Germany Proofs received from author(s): February 19, 2007